pg.ddx.io  pgsql-bugs@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Ayush Tiwari <ayushtiwari.slg01@gmail.com>
Cc: Борис Мурашов <murasovb@gmail.com>
Cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #19906: Backend crash (signal 11) creating a LANGUAGE sql function that returns a composite type from a sub-
Date: Tue, 17 Nov 2026 11:34:18 -0500
Message-ID: <457447.1794933258@sss.pgh.pa.us> (raw)
In-Reply-To: <CAJTYsWU_eHpULkHVOpAMz39y=96hffgud4nExWpj0xL6cEgt5Q@mail.gmail.com>
References: <19906-229033275b593ef0@postgresql.org>
	<CAJTYsWUm1j=YQsHpt12=E_Xmx7GhtmVtRAZ2x+ySOo3bkJtWyQ@mail.gmail.com>
	<126458.1794857339@sss.pgh.pa.us>
	<CAF_qW9dJpZV0Cv_F3rUh=77dAtQGbMjjEf9v7ng4mz3Ai=cngQ@mail.gmail.com>
	<266210.1794869177@sss.pgh.pa.us>
	<CAJTYsWU_eHpULkHVOpAMz39y=96hffgud4nExWpj0xL6cEgt5Q@mail.gmail.com>

Ayush Tiwari <ayushtiwari.slg01@gmail.com> writes:
> Post your last mail, I was auditing the places that need these comment
> updates, and the difference between that list and your patch is:
> 1) make_fn_arguments() in parse_func.c still has the old comment, and
>    clauses.c calls it with NULL.
> 2) make_row_comparison_op() in parse_expr.c has it too, but it
>    dereferences pstate anyway, so that line can just go?

Good catches.  Incorporated in less-drafty v2.

			regards, tom lane

Attachments:

v2-0001-Don-t-crash-on-null-pstate-in-coerce_record_to_co.patch (text/x-diff, 9.2K) download From 308386876e33e0a7569c53372af1050cb4a5d8bc Mon Sep 17 00:00:00 2001 From: Tom Lane <tgl@sss.pgh.pa.us> Date: Tue, 17 Nov 2026 11:32:47 -0500 Subject: [PATCH v2] Don't crash on null pstate in coerce_record_to_complex(). Since commit 913bbd88d, this case is reachable if a SQL-language function returns a whole-row Var of type RECORD when its declared output type is a named composite type. We'd need the original ParseState to find out what the Var's columns are, but we don't have that structure at hand anymore. This didn't work before 913bbd88d either, but you got a slightly off-point error message instead of a crash. Given that this has never worked and there's been no field complaints about that before now, it doesn't seem worth doing major work to fix this; let's just ensure that we throw a helpful error explaining what to do to work around the limitation. (The workaround actually reaches this same code, but does so while we still have the ParseState.) While at it, update various badly-obsolete comments about what coerce_type() uses its pstate argument for. Bug: #19906 Reported-by: Boris Murasov <murasovb@gmail.com> Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Ayush Tiwari <ayushtiwari.slg01@gmail.com> Discussion: https://postgr.es/m/19906-229033275b593ef0@postgresql.org Backpatch-through: 15 --- src/backend/parser/parse_coerce.c | 46 +++++++++++++++--------- src/backend/parser/parse_expr.c | 4 +-- src/backend/parser/parse_func.c | 3 +- src/test/regress/expected/rangefuncs.out | 9 ++++- src/test/regress/sql/rangefuncs.sql | 7 +++- 5 files changed, 45 insertions(+), 24 deletions(-) diff --git a/src/backend/parser/parse_coerce.c b/src/backend/parser/parse_coerce.c index d3240f4b265..bd4e12c9989 100644 --- a/src/backend/parser/parse_coerce.c +++ b/src/backend/parser/parse_coerce.c @@ -147,9 +147,14 @@ coerce_to_target_type(ParseState *pstate, Node *expr, Oid exprtype, * applies length conversion, and in those cases only, the result will * already be properly coerced to the specified typmod. * - * pstate is only used in the case that we are able to resolve the type of - * a previously UNKNOWN Param. It is okay to pass pstate = NULL if the - * caller does not want type information updated for Params. + * pstate can be NULL in calls from outside the parser. This mostly works, + * but note the following restrictions: + * 1. No p_coerce_param_hook processing is possible, so Params receive + * no special treatment. + * 2. Whole-row Vars of type RECORD (those referring to non-table FROM items + * such as joins or sub-selects) cannot be cast to named composite types, + * since we can't resolve their columns. + * 3. No error position information will be provided. * * Note: this function must not modify the given expression tree, only add * decoration on top of it. See transformSetOperationTree, for example. @@ -1009,7 +1014,7 @@ build_coercion_expression(Node *node, * Coerce a RECORD to a specific composite type. * * Currently we only support this for inputs that are RowExprs or whole-row - * Vars. + * Vars. pstate == NULL prevents handling the latter case. */ static Node * coerce_record_to_complex(ParseState *pstate, Node *node, @@ -1039,12 +1044,22 @@ coerce_record_to_complex(ParseState *pstate, Node *node, else if (node && IsA(node, Var) && ((Var *) node)->varattno == InvalidAttrNumber) { - Var *var = (Var *) node; - ParseNamespaceItem *nsitem; + if (pstate) + { + Var *var = (Var *) node; + ParseNamespaceItem *nsitem; - nsitem = GetNSItemByVar(pstate, var); - args = expandNSItemVars(pstate, nsitem, var->varlevelsup, - var->location, NULL); + nsitem = GetNSItemByVar(pstate, var); + args = expandNSItemVars(pstate, nsitem, var->varlevelsup, + var->location, NULL); + } + else + ereport(ERROR, + (errcode(ERRCODE_CANNOT_COERCE), + errmsg("cannot cast type %s to %s", + format_type_be(RECORDOID), + format_type_be(targetTypeId)), + errhint("Cast the whole-row variable to the desired type explicitly."))); } else ereport(ERROR, @@ -1156,8 +1171,7 @@ coerce_record_to_complex(ParseState *pstate, Node *node, * * Returns the possibly-transformed node tree. * - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * As with coerce_type, pstate may be NULL for calls from outside the parser. */ Node * coerce_to_boolean(ParseState *pstate, Node *node, @@ -1203,8 +1217,7 @@ coerce_to_boolean(ParseState *pstate, Node *node, * * Returns the possibly-transformed node tree. * - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * As with coerce_type, pstate may be NULL for calls from outside the parser. */ Node * coerce_to_specific_type_typmod(ParseState *pstate, Node *node, @@ -1252,8 +1265,7 @@ coerce_to_specific_type_typmod(ParseState *pstate, Node *node, * * Returns the possibly-transformed node tree. * - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * As with coerce_type, pstate may be NULL for calls from outside the parser. */ Node * coerce_to_specific_type(ParseState *pstate, Node *node, @@ -1330,6 +1342,7 @@ parser_coercion_errposition(ParseState *pstate, * This is used for determining the output type of CASE, UNION, * and similar constructs. * + * 'pstate' is used only for error reporting, and can be NULL. * 'exprs' is a *nonempty* list of expressions. Note that earlier items * in the list will be preferred if there is doubt. * 'context' is a phrase to use in the error message if we fail to select @@ -1569,8 +1582,7 @@ select_common_type_from_oids(int nargs, const Oid *typeids, bool noerror) * expressions to the desired type. 'context' is a phrase to use in the * error message if we fail to coerce. * - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * As with coerce_type, pstate may be NULL for calls from outside the parser. */ Node * coerce_to_common_type(ParseState *pstate, Node *node, diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c index 728392f436f..e207b7a1978 100644 --- a/src/backend/parser/parse_expr.c +++ b/src/backend/parser/parse_expr.c @@ -2826,9 +2826,7 @@ transformCollateClause(ParseState *pstate, CollateClause *c) /* * Transform a "row compare-op row" construct * - * The inputs are lists of already-transformed expressions. - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * The argument lists are lists of already-transformed expressions. * * The output may be a single OpExpr, an AND or OR combination of OpExprs, * or a RowCompareExpr. In all cases it is guaranteed to return boolean. diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c index 410e54accd8..c4ba284fa18 100644 --- a/src/backend/parser/parse_func.c +++ b/src/backend/parser/parse_func.c @@ -2005,8 +2005,7 @@ unify_hypothetical_args(ParseState *pstate, * * Caution: given argument list is modified in-place. * - * As with coerce_type, pstate may be NULL if no special unknown-Param - * processing is wanted. + * As with coerce_type, pstate may be NULL for calls from outside the parser. */ void make_fn_arguments(ParseState *pstate, diff --git a/src/test/regress/expected/rangefuncs.out b/src/test/regress/expected/rangefuncs.out index a7cb1b5611d..a9ad3687887 100644 --- a/src/test/regress/expected/rangefuncs.out +++ b/src/test/regress/expected/rangefuncs.out @@ -2,7 +2,14 @@ CREATE TABLE rngfunc2(rngfuncid int, f2 int); INSERT INTO rngfunc2 VALUES(1, 11); INSERT INTO rngfunc2 VALUES(2, 22); INSERT INTO rngfunc2 VALUES(1, 111); -CREATE FUNCTION rngfunct(int) returns setof rngfunc2 as 'SELECT * FROM rngfunc2 WHERE rngfuncid = $1 ORDER BY f2;' LANGUAGE SQL; +CREATE FUNCTION rngfunct(int) returns setof rngfunc2 as + 'SELECT * FROM rngfunc2 WHERE rngfuncid = $1 ORDER BY f2;' LANGUAGE SQL; +-- this variant doesn't work +CREATE FUNCTION rngfunct2(int) returns setof rngfunc2 as + 'SELECT s FROM (SELECT * FROM rngfunc2) s;' LANGUAGE SQL; +ERROR: cannot cast type record to rngfunc2 +HINT: Cast the whole-row variable to the desired type explicitly. +CONTEXT: SQL function "rngfunct2" -- function with ORDINALITY select * from rngfunct(1) with ordinality as z(a,b,ord); a | b | ord diff --git a/src/test/regress/sql/rangefuncs.sql b/src/test/regress/sql/rangefuncs.sql index 3c47c98e113..1d35f938fd4 100644 --- a/src/test/regress/sql/rangefuncs.sql +++ b/src/test/regress/sql/rangefuncs.sql @@ -3,7 +3,12 @@ INSERT INTO rngfunc2 VALUES(1, 11); INSERT INTO rngfunc2 VALUES(2, 22); INSERT INTO rngfunc2 VALUES(1, 111); -CREATE FUNCTION rngfunct(int) returns setof rngfunc2 as 'SELECT * FROM rngfunc2 WHERE rngfuncid = $1 ORDER BY f2;' LANGUAGE SQL; +CREATE FUNCTION rngfunct(int) returns setof rngfunc2 as + 'SELECT * FROM rngfunc2 WHERE rngfuncid = $1 ORDER BY f2;' LANGUAGE SQL; + +-- this variant doesn't work +CREATE FUNCTION rngfunct2(int) returns setof rngfunc2 as + 'SELECT s FROM (SELECT * FROM rngfunc2) s;' LANGUAGE SQL; -- function with ORDINALITY select * from rngfunct(1) with ordinality as z(a,b,ord); -- 2.52.0

view thread (9+ messages)  latest in thread

Message-ID: <457447.1794933258@sss.pgh.pa.us>
Permalink:  ../457447.1794933258@sss.pgh.pa.us/
Also on:    postgresql.org/message-id/457447.1794933258@sss.pgh.pa.us

 ·  · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients from your mail client; the link fills in
  the headers below: reply via email
  Some clients, webmail especially, drop In-Reply-To and References from
  a link, and the reply then starts a new thread: use the mbox below.

  To: pgsql-bugs@postgresql.org
  Cc: tgl@sss.pgh.pa.us, ayushtiwari.slg01@gmail.com, murasovb@gmail.com
  Subject: Re: BUG #19906: Backend crash (signal 11) creating a LANGUAGE sql function that returns a composite type from a sub-
  In-Reply-To: <457447.1794933258@sss.pgh.pa.us>
  References: <19906-229033275b593ef0@postgresql.org> <CAJTYsWUm1j=YQsHpt12=E_Xmx7GhtmVtRAZ2x+ySOo3bkJtWyQ@mail.gmail.com> <126458.1794857339@sss.pgh.pa.us> <CAF_qW9dJpZV0Cv_F3rUh=77dAtQGbMjjEf9v7ng4mz3Ai=cngQ@mail.gmail.com> <266210.1794869177@sss.pgh.pa.us> <CAJTYsWU_eHpULkHVOpAMz39y=96hffgud4nExWpj0xL6cEgt5Q@mail.gmail.com> <457447.1794933258@sss.pgh.pa.us>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox